Skip to content

Conversation

@bergey
Copy link

@bergey bergey commented Feb 17, 2021

The monotonic clock appears to be faster to query, as well as alledgedly more accurate. Closes #13.

Running the benchmark on my Macbook, I see 10-30 microseconds for the version in this PR, vs ~130 microseconds for the version in master.

@bergey
Copy link
Author

bergey commented Feb 18, 2021

I don't understand the build failure. I want to believe that it is not due to changes in this PR (since I did not change any types) but the last build of master succeeded.

@arianvp
Copy link
Member

arianvp commented Jan 24, 2022

Master branch seems to not compile the tests. #29 . We'll probably fix in #28 . It's not related to your change

@arianvp
Copy link
Member

arianvp commented Jan 24, 2022

This change looks good overall. Once we have the main branch compiling again we can merge this

@arianvp arianvp self-requested a review January 24, 2022 15:08
@arianvp
Copy link
Member

arianvp commented Feb 14, 2022

@bergey if you rebase you should trigger CI again and this can be merged

@bergey bergey force-pushed the monotonic-ghc.clock branch from e05c471 to d9387b6 Compare March 30, 2022 23:58
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

use monotonic clock instead of system clock

2 participants